Deep Learning: Facebook Wants To Know Even More About Its Users | Hacker News Reader

Deep Learning: Facebook Wants To Know Even More About Its Users

allfacebook.com

52 pointsdave_sullivan26 comments

This article adds nothing except sensationalism to the original article by MIT Technology Review http://www.technologyreview.com/news/519411/facebook-launche...

"Facebook Launches Advanced AI Effort to Find Meaning in Your Posts"

Funny how the headline looks as if facebook couldn't find any meaning in user posts, and they are now desperately turning to AI for help.

The headline amuses me. Of course they want to know even more about their users! What data-driven company wouldn't? This is like saying "Apple wants to make even more money" or "GM wants to sell even more cars."

FWIW, my original headline was "Facebook interested in deep neural networks" but it was apparently modded to the headline of the blog post. I consider the modded title more sensationalist (and obvious), but hey...

I presumed this is going to be about Facebook's constant push for users to fill out more "profile information ". Every time I go to my profile, I am presented with a banner about updating my info… They ask for more school details, or lately, books, music and movies that I'm familiar with. One day I filled out something about movies at their request, then I saw on my girlfriends newsfeed and I've been saying that I added movies to the list of movies I'd watched! Which I suppose I had but it seemed somewhat disingenuous since it was at their urging.

Yeah, it's the same on LinkedIn and Google+ lately (on LinkedIn you get it reduced to those extra ridiculous endorsements...just a binary "do you think person X is skilled at typing using a keyboard?"). They've learned that if you ask single questions people actually answer them, while many people find an entire empty profile daunting and so don't get far (or never even start).

Completely off topic, but the Facebook movie thing is weird. They no longer ask for your favorite movies in your profile, which at least somewhat lines up with a "profile" of a person, as your favorites say something about you (true or not). Now your profile just asks for "movies you've watched", which is bizarre. Most people have seen a lot of movies; what helpful information does that communicate to Facebook or to your Facebook friends, which is supposed to be the point? Certainly it's unlikely that you'll be able to fill that out comprehensively, or with a sampling good enough that some ML algorithm can make something of it. I guess it is self selected, which says something about what movies you've seen, can recall, and are willing to disclose, but it still seems to be setting about the lowest bar possible. Is it going to ask for "sports you've watched" next? "Celebrities you've heard of"?

Edit: maybe it's just another user psychology thing. "Favorite movies" might make people worry too much about what to put, while "movies you've watched" is meant to be a super low bar, and people will still tend to put their favorite movies in anyway? Those people who love updating their profiles will dilute that signal by adding every new movie they see going forward, but that's just a different, still useful signal.

Part of why I felt it out is that it was presented to me as very low-key thing, just finishing profile information… "Movies I've watched", which as you said is very casual. But then it was presented to others like information that I (I'm grandfathered into a grooveshark subscription at $3 a mo. and have kept it despite the fact that a I have a spotify subscription...)specifically felt the need to go add to my profile on my own, as if it was important that I watched some obscure movie from 1982 and really wanted my friends to know. It's even sillier that they were movies from a random FB list of 3 choices for each question, not something I typed in.

One could argue that's the #1 priority they should be focused on. One that any advertising company would consider very critical. But I guess use any emerging buzzword like "deep learning", and people will find it news-worthy.

I think this would be fascinating to work on, the size and variety of FB's data is mind boggling. If I was looking for a job and had more ML experience (it's only been a focus of mine for the past year or so) then I would be finding out how to get on that team.

Any interesting ideas I have involve a heavy machine learning component. I'm sure the most valuable software companies in the future will have too.

You might try NSA, they have possibly even more data to boggle ones mind. You find that fascinating because of the means or of the ends of it?

Ha, unfortunately:

- I'm not American

- I'd rather make stuff that improves people's lives in any way, and not subvert their freedoms.

ML is fascinating to me because what can be done with it.

Better not join Facebook then

Ah yes. There are still some little shreds of privacy left somewhere in the world, so why not help Facebook with that? It's an engineering challenge, Yay!

Also, Wall Street is not yet completely done destroying the US economy (and others with it), and I bet they could use your help too. You could do your part in providing liquidity to the markets. Everyone loves liquidity! Besides, it's challenging, mathy work! Yay.

.. Disgusting.

Same here. I've also been heavily occupied with ML and AI for the past year or two. I think it's going to usher in the next phase of computers, we'll see. Computers are finally getting fast enough to do some interesting things with all of this data we're creating.

Google hires Prof Hinton. Facebook counters. There was a subtle point in the article referencing the fact that this challenge has both a software and hardware component. Add in complexities of parallel processing and running on GPUs and you have a pretty limited talent poop to pick from.

Glad to see the interest in deep learning even if it is sensationalizing it a bit. More interest and a bigger community will be helpful for startups like us.

An interesting thing to note: Facebook actually hired Marc’Aurelio Ranzato away from Google, where he was working on the Brain project. Prior to that he was working in Hinton's lab at University of Toronto.

http://www.cs.toronto.edu/~ranzato/

Facebook Recruits on Kaggle, a Machine Learning competition site: http://www.kaggle.com/c/facebook-recruiting-iii-keyword-extr...

I love it. "Do some free work for us and maybe we'll hire you!"

It's not for them though. It's just data science competitions that anyone can enter.

Yet more money to be made selling our thoughts then.

Don't sell your thoughts to Facebook then. Very simple.

> sell

Worse, people freely give them away.

To help users or advertisers?

Why even ask...

I would bet both.

Facebook without users don't have anyone to advertise to. But obviously they want to be efficient to advertisers, too...

I'm curious what kind of data they want to analyze with deep learning.

Reply on news.ycombinator.com